I have a project where it is very tempting to use a deeply nested collection of data. But how does that affect performance? Are there any reasons why I should avoid doing this? I'm talking about lists in lists in lists, etc.
Per this StackOverflow thread, it sounds like lists are created by serializing and then deserializing the data.
If you were deeply nesting your documents with list inside of list, that would be a serious performance hit purely from the serialization and deserialization.
If you have a hierarchy like this, I would consider using MySQL, since it's designed for relational data. It would be simple to implement a hierarchy using a relational DBMS compared to CouchDB.